Invariant Event Tracking on Social Networks

Abstract

When an event is emerging and actively discussed on social networks, its related issues may change from time to time. People may focus on different issues of an event at different times. An invariant event is an event with changing subsequent issues that last for a period of time. Examples of invariant events include government elections, natural disasters, and breaking news. This paper describes our demonstration system for tracking invariant events over social networks. Our system is able to summarize continuous invariant events and track their developments along a timeline. We propose invariant event detection by utilizing an approach of Clique Percolation Method (CPM) community mining. We also present an approach to event tracking based on the relationships between communities. The Twitter messages related to the 2013 Australian Federal Election are used to demonstrate the effectiveness of our approach. As the first of this kind, our system provides a benchmark for further development of monitoring tools for social events.